An AI performance audit for AI that isn’t paying off

For companies whose AI is live but not hitting its goal, or about to go live and needing an independent test first. You get one clear answer: keep, fix, replace or stop.

Report done in 2 to 6 weeks
Book a free intro call

A free 15 minute intro call with our managing director.

56%of CEOs have seen no significant financial benefit from AI. We find out why.PwC Global CEO Survey, January 2026, 4,454 CEOs worldwide

Six links, shown on an example

  1. 1

    Goal

    What the AI is meant to improve, written as a number.

    Holds
  2. 2

    Design

    How the tool fits the work and the people who use it.

    Holds
  3. 3

    Build

    The tool is built and set up the way it was designed.

    First break found
  4. 4

    Data

    The right data reaches it, clean and connected.

    Fix 3 first
  5. 5

    Use

    People actually use it in their daily work.

    Fix 3 first
  6. 6

    Measurement

    Results are tracked against where you started.

    Fix 3 first

We test six links in order and stop at the first one that broke.

What we audit

One tool, or a custom agent or system. Both end in a clear answer and a costed plan.

An AI tool

For one off-the-shelf tool, such as a chatbot, copilot or AI add-on.

Sounds like you if:

  • “Our chatbot went live, and the support tickets didn’t go down.”
  • “People tried the tool once and went back to the old way.”
  • “A renewal is coming, and I can’t tell if it’s earning its keep.”

One answer after the report: keep, fix, replace or stop.

Report done in 2 to 3 weeks

A custom agent or system

For an agent, bot or automation a vendor, agency or your team built.

Sounds like you if:

  • “Our AI went live months ago and still hasn’t hit its goal.”
  • “The vendor says it’s our data. We say it’s their build.”
  • “We’re about to launch an AI agent and want it tested first.”

Root causes, a costed fix, and the builder’s written reply.

Report done in 3 to 6 weeks

Why our verdict holds up

It comes from your systems, from people with nothing to sell, and the builder gets a say.

Evidence from your systems

Usage logs, output quality tests, your data and integrations, checked against how your team really works.

We stay independent

We accept no vendor money and have no bias toward any vendor. Our duty is to our clients.

A fair hearing

With your OK, whoever built it checks our findings in writing within 5 business days, before anything is final.

Fixed fee, fixed dates

Quoted in writing after a briefing. No hourly billing.

What you get

Your audit report

For each tool or system we audit

  • A clear answer

    Keep, fix, replace or stop, with the reasons behind it.

  • The root causes

    What is holding it back, and how strong the evidence is behind each cause.

  • A costed plan

    What to do next and what it costs, written so any developer can follow it.

  • A report and a live presentation

    Written findings, then a walk through with your leadership team.

Your options

From $9,500 to $47,500, by scope. Your fixed fee, scope and dates come in writing after a briefing. Payment in stages: under $10,000 at signing; $10,000 to $50,000 half at signing and half at the readout.

Audit options

  • Tool audit

    One off the shelf AI tool

    2 to 3 weeks
  • System audit

    One custom agent, bot or automation

    3 to 4 weeks
  • Multisystem audit

    Up to 3 systems, or everything from one vendor

    5 to 6 weeks

Report done in the time shown. Bars sit on a 0 to 8 week scale.

The fine print

  • Fixed fees and dates in writing. No hourly billing.
  • We agree the targets and how to measure them before we look at results.
  • We take no money from any AI vendor or developer, and we tell you if a partner introduced us.
  • Defects in the original build stay with the builder.

Questions we hear

Is this an AI audit?

Yes. It’s an independent AI performance audit: we test whether a live AI tool or system does what it was bought or built to do, and why not.

What if our AI is working fine?

Then the audit says keep, with the numbers to show it: who uses it and what it costs per active user.

Can you fix or rebuild it for us?

No. We plan it, audit it or take what’s built live. Defects stay with the builder, and our plan goes to any developer you choose.

Does whoever built it get a say?

Yes. With your OK, they check our findings in writing within 5 business days, before anything is final.

Can you test an AI agent before it goes live?

Yes. We test it against written acceptance criteria: accuracy, failure cases, data access, security and cost per task.

Audit or assessment: which do we need?

An AI Performance Audit goes deep on one tool or system. An AI Readiness Assessment looks across everything you pay for. The free intro call sorts it out.

That’s every detail. Questions left? Ask us in person.

Get one clear answer
about your AI

Start with a free 15 minute intro call with our managing director. If it’s a fit, we book a 60 minute briefing with your leadership for another day. Within 7 business days of it, you get up to three fixed fee options, ours marked, and ‘not yet’ if that’s the honest answer.